Fraser et al. (2016) : Phonation rate: "the proportion of an utterance that is vocalized; low values here refer to more time being spent silently, as in a pause.”
Beltrami et al. (2016): Silence segments duration: mean, median and Std. Dev significant features in all tasks. They also calculate the same for speech, but I don't think we need both since will be highly correlated?
Possible implementation
References (AD/MCI):